Who Owns AI-Assisted Research Code or Designs?

If AI Helps Write Your Research Code or Create a Design, Who Owns the Result?

You are developing software for a research project and use generative artificial intelligence to write functions, debug algorithms, or construct a data-processing pipeline. Perhaps you use an AI image generator to develop a scientific illustration, prototype interface, or engineering design.

The resulting work becomes an important part of your research. You may want to publish the code, deposit it in a repository, license it to other researchers, or commercialize the technology.

But who owns what AI helped create? Do you automatically hold the copyright because you supplied the prompts? Could your university claim the work? Does the AI provider retain rights, and can you legally distribute the generated material?

Answering these questions requires distinguishing ownership from copyright protection, contractual permissions, and the rights of third parties.

AI Assistance Does Not Automatically Establish Ownership

How Intellectual-Property Rights Apply to AI-Assisted Research Code and Designs

Research code and designs can involve different intellectual-property rights

Research software and technical designs may be protected through different legal mechanisms, depending on what has been created and the jurisdiction involved.

Copyright may protect original computer code, technical drawings, illustrations, and other forms of expression. Patent law may protect qualifying technical inventions. Trade secret law may protect certain confidential information, while contracts may govern access, use, ownership, and distribution.

These rights are not interchangeable.

For example, a researcher may hold rights in original source code without having exclusive rights to the mathematical method implemented by that code.

Similarly, owning a technical drawing does not automatically establish ownership of a patent covering the device depicted.

Can AI-generated research code receive copyright protection?

Potentially, but the answer depends on human authorship and the applicable jurisdiction.

In the United States, the Copyright Office has maintained that copyright protection requires human authorship. Its January 2025 report on copyrightability concluded that generative AI outputs may receive protection where a human author determines sufficient expressive elements.

Human-authored material incorporated into an AI-assisted work may be protected. Creative human modifications or arrangements may also qualify, depending on their nature.

However, merely supplying prompts does not necessarily establish copyright authorship of the resulting output.

This distinction matters for research code generated largely through conversational instructions.

A researcher who independently develops original program structure, writes substantial portions of the implementation, and creatively revises AI suggestions may have protectable human-authored contributions.

By contrast, a researcher who accepts a complete AI-generated script without meaningful human authorship should not automatically assume exclusive copyright protection over the generated expression.

The United States Copyright Office's position is jurisdiction-specific. Other countries may apply different statutory approaches to computer-generated works.

Does writing the prompt make you the copyright owner?

Not automatically.

A prompt may express the researcher's intentions, specifications, or desired functionality. However, specifying the outcome is not necessarily equivalent to authoring the expressive elements of the resulting code or design.

For example, asking an AI system to create a Python script that cleans survey data establishes the requested functionality. It does not necessarily mean the researcher determined the particular source-code expression produced by the system.

Similarly, requesting a scientific illustration showing a laboratory instrument does not automatically establish authorship of every visual element in the generated image.

The analysis becomes more nuanced when the researcher exercises substantial creative control, contributes original material, or meaningfully modifies the output.

Human authorship should be evaluated through the actual creative contribution rather than the number or complexity of prompts alone.

What if you substantially modify AI-generated code?

Human modifications may qualify for copyright protection if they contain sufficient original expression under the applicable law.

Suppose an AI system generates a preliminary data-analysis script. The researcher then restructures the program, writes original functions, develops a distinctive software architecture, and revises the implementation to support a particular research methodology.

Those human-authored contributions may be protectable even if some AI-generated material remains in the final program.

However, copyright protection does not necessarily extend to every element of the combined work.

Common programming patterns, functional requirements, mathematical principles, and other unprotectable elements may remain outside copyright protection.

The relevant question is not simply how many lines the researcher changed. It is whether the human contribution contains qualifying original expression.

Does the AI provider own the generated code or design?

Not automatically. The answer depends partly on the applicable service agreement.

Some AI providers state that they do not claim ownership of generated outputs or assign whatever rights they may have to users. However, these contractual provisions do not necessarily establish that the output is copyrightable or that the user possesses exclusive rights against everyone else.

GitHub, for example, states in its Copilot documentation that it does not claim ownership of suggestions. It also acknowledges that whether a suggestion can be owned depends on factors such as applicable law and the nature of the generated material.

The provider's position should not be confused with a legal determination that every suggestion is an original, copyright-protected work.

Researchers should also examine whether service terms permit commercial use, impose restrictions, or grant the provider licenses relating to submitted inputs and generated outputs.

These arrangements may differ between consumer services, institutional subscriptions, and enterprise agreements.

Can your university own AI-assisted research code?

Yes. A university may hold rights in research software under applicable law, employment arrangements, intellectual-property policies, sponsored research agreements, or assignments.

However, university ownership is not universal.

Different rules may apply to faculty research software, student projects, employee-developed systems, independently created software, and work produced under external funding.

For example, software developed as part of assigned employment duties may be subject to different ownership arrangements from software created independently outside those duties.

The use of university equipment or an institutional AI subscription may be relevant under particular agreements, but neither fact automatically determines ownership in every jurisdiction.

Researchers should review the applicable documents rather than relying on assumptions about academic freedom or the identity of the person who wrote the prompts.

Where the software implements a potentially patentable technical invention, ownership of the research invention may require an additional analysis separate from copyright in the source code.

Can AI-generated code infringe someone else's copyright?

Potentially. AI-generated code may resemble existing copyrighted software or incorporate material subject to third-party licensing conditions.

Similarity does not automatically establish infringement. The legal analysis depends on the nature and amount of protected expression, applicable exceptions, licensing arrangements, and other circumstances.

Nevertheless, researchers should not assume that generated code is free of intellectual-property obligations simply because it was produced by an AI system.

GitHub's Copilot documentation recognizes that generated suggestions may resemble existing code and provides mechanisms for identifying certain matching public-code references.

These mechanisms can support review, but they do not guarantee that all relevant third-party rights will be detected.

For research software intended for publication or commercialization, code review should include appropriate license-compliance checks alongside security and functional testing.

What happens when AI-generated code resembles open-source software?

Open-source software is not necessarily free from copyright restrictions. Its licenses grant permissions subject to particular terms.

Some licenses require preservation of copyright notices or attribution. Others impose conditions concerning redistribution, source-code availability, or derivative works.

If generated code reproduces protectable material from an existing project, the relevant license and legal obligations may need to be considered.

However, the mere resemblance of a short functional code fragment to an open-source implementation does not automatically establish that a license applies.

Researchers should assess the nature of the similarity and the applicable legal requirements.

When substantial matching code is identified, the safest practical response may be to investigate its provenance, examine the relevant license, and seek appropriate guidance before redistribution.

Can you release AI-assisted research code under an open-source license?

Often, yes, provided that you have the necessary rights and authority to license the material.

However, applying an open-source license does not create copyright ownership over material that is not legally protectable or that belongs to someone else.

A researcher may be able to license qualifying human-authored portions of a program while other elements remain subject to different rights or restrictions.

Institutional ownership policies may also require authorization before public release.

For example, a researcher cannot necessarily place university-owned software under a permissive license simply because the code repository is maintained through a personal account.

Before release, verify the rights holder, applicable third-party licenses, and any sponsor or institutional obligations.

Are AI-generated engineering designs treated the same as AI-generated code?

Not necessarily. Technical designs may involve copyright, patent rights, registered industrial designs, trade secrets, or combinations of these protections.

A two-dimensional engineering drawing may contain copyright-protected expression, while the functional device depicted may raise separate patent questions.

A graphical user interface may contain protectable visual expression without necessarily giving its creator exclusive rights to the underlying software functionality.

Similarly, an AI-generated product shape may require assessment under industrial-design law rather than copyright alone.

The relevant legal framework depends on what is being protected.

Researchers should distinguish the visual or textual representation of a design from the underlying technical invention.

The separate question of patentability of an AI-assisted technical invention becomes relevant when the design embodies a novel functional solution.

Does AI-assisted code ownership affect reproducibility or publication?

Potentially. Intellectual-property restrictions may determine whether researchers can publish source code, distribute executable software, or provide access to proprietary implementations.

A journal or funder may encourage code sharing, but researchers must still consider applicable licenses, confidentiality obligations, and institutional policies.

Where public release is restricted, alternative reproducibility arrangements may sometimes be possible, such as controlled access, detailed methodological documentation, or independently developed implementations.

These alternatives should not be presented as equivalent to unrestricted code availability.

Researchers should also document relevant AI assistance when required by the journal, institution, or funder. Such documentation supports transparency but does not itself resolve copyright ownership.

A Researcher Uses AI to Develop a Data-Analysis Application

The use of AI does not automatically prevent the researcher from releasing the application. However, the researcher must distinguish the right to use generated suggestions from the authority to license the complete software project.

The outcome may differ if the application was developed under a sponsored research agreement or contains proprietary institutional software.

Common Misconceptions About AI-Generated Code and Design Ownership

Establish Your Rights Before Publishing, Licensing, or Commercializing AI-Assisted Work

The practical starting point is to identify what you want to do with the code or design.

Using software internally for research may raise different questions from distributing it publicly, selling licenses, or incorporating it into a commercial product.

Ownership, permission to use, and permission to redistribute should be assessed separately.

Researchers should also consider whether the work contains unpublished technical information that should not be submitted to external AI services.

The confidentiality of unpublished research inventions can be relevant even when the immediate task involves writing code or generating a design.

For commercially sensitive projects, institutional restrictions on AI use may determine which tools and deployments are authorized.

Before Releasing or Commercializing AI-Assisted Research Code or Designs

Frequently Asked Questions About AI-Assisted Code and Design Ownership

Using AI Does Not Automatically Give You Exclusive Rights to Research Code or Designs

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.